/cloudbiolinux

CloudBioLinux: configure virtual (or real) machines with tools for biological analyses

Primary LanguagePythonMIT LicenseMIT

CloudBioLinux is a build and deployment system which installs an easily customizable selection of bioinformatics and machine learning libraries on a linux container, bare virtual machine (VM) image, freshly installed PC, or in the cloud. CloudBioLinux is a curated and community developed set of instructions for tools provided by operating system packages (debs and RPMs), external packaging efforts (bioconda and homebrew-science) and language specific library installers (Python, R, Perl and Ruby).

CloudBioLinux included software packages are fully customizable. In addition to the default configuration, we support custom configuration builds through flavors. Flavors support overriding default package installations, making it simple to create derived installs for specific purposes.

CloudBioLinux is a single install route for Docker containers ,desktop VMs such as VirtualBox, cloud providers such as Amazon EC2 or desktop machines. This works equally well for other virtual machines and private cloud environments, including XEN, Linux KVM, Eucalyptus and Openstack.

Quick start

bcbio uses CloudBioLinux as the basis for tool installation and provides a large set of supported and tested tools. If you're looking to bootstrap a system with tools and data for high throughput sequencing analysis, we suggest using the bcbio installer which fully wraps CloudBioLinux and provides an easy path to customize install directories, organisms and biological data installed.

We recommend using, or developing, a custom flavor to choose tools of interest to install. The amount of bioinformatics software continues to increase -- there are over 1000 recipes in bioconda -- and it's difficult to come up with a default installation that includes everything for everyone. The ngs_pipeline_minimal flavor has the set of NGS analysis tools installed with bcbio and is a good starting point for understanding the CloudBioLinux install process:

fab -f fabfile.py -H localhost install_biolinux:flavor=ngs_pipeline_minimal

Installing CloudBioLinux on a local machine

The install process for CloudBioLinux is fully automated through a Fabric build file written in Python. Everything is fully configurable through plain text YAML configuration files, and custom build targets allow installation of a subset of the total available packages.

Setup

Retrieve the CloudBioLinux code base and install fabric:

pip install fabric
git clone git://github.com/chapmanb/cloudbiolinux.git
cd cloudbiolinux

Usage

The basic usage specifies the hostname of a machine accessible via ssh or the local machine:

fab -f fabfile.py -H localhost install_biolinux

Fabric contains some other useful commandline arguments for customizing this to your environments:

  • -c your_fabricrc.txt -- Specify the path to a fabricrc configuration files. This allows customization of install directories and other server specific details. See the default config/fabricrc.txt for a full list of options.
  • -u username -- The username on a remote machine, overriding the default of your current username.

Customization with flavors

In most cases you want to customize a specific set of packages, or install into an isolated directory without root access, using flavors:

fab -f fabfile.py -H localhost install_biolinux:flavor=my_flavor

my_flavor can be the name of an existing flavor in contrib/flavor or the path to a directory with customization information. The files in your flavor directory replace those in the standard config directory, allowing replacement of any of the configuration files like main.yaml with customized copies. If you desire even more control, flavors allow custom python hooks. See doc/hacking.md for more details.

The best place to get started is the demo flavor included with CloudBioLinux. This installs a small number of common packages into an isolated directory (~/tmp/cbl_demo by default), without root access. Run the example with:

fab -f fabfile.py -H localhost install_biolinux:flavor=demo

Specific install targets

You can substitute install_biolinux with more specific targets to only build portions of CloudBioLinux:

  • install_biolinux:packages -- Install all of the defined system packages.
  • install_biolinux:libraries -- Install all libraries for various programming languages.
  • install_biolinux:brew -- Install homebrew packages only.
  • install_libraries:language -- Install libraries for a specific language.
  • install_biolinux:custom -- Install all custom programs.
  • install_brew:a_package_name -- Install a specific brew package.
  • install_custom:a_package_name -- Install a specific custom program.

Homebrew package installation

Homebrew and Linuxbrew provide a Ruby-based environment for installing packages on MacOSX and Linux. The active homebrew-science packaging community maintains a number of common scientific tools. We also maintain a homebrew-cbl repository with tools not yet integrated into homebrew-science.

CloudBioLinux manages installation of the Linuxbrew or Homebrew framework and pulls in the homebrew/science and chapmanb/cbl taps, as well as injecting your current compilers into the homebrew build scripts. To install a supported package using CloudBioLinux:

fab -f fabfile.py -H localhost install_custom:bedtools

Specific package installation

The custom directory contains installation instructions for programs that are not available from standard package repositories, written in Python using the Fabric remote deployment tool. To install individual custom packages:

fab -f fabfile.py -H localhost install_custom:your_package_name

We prefer using the Homebrew framework for new packages over writing custom packages.

Biological data

We manage a repository of useful public biological data on an Amazon S3 bucket. Currently this includes whole genomes pre-indexed for a number of popular aligners. Downloading and installing these saves a ton of time over running the indexing steps yourself, and eases running next-generation analyses on cloud machines.

A Fabric build script is provided to install this data on your local machine. A biodata configuration file in YAML format, config/biodata.yaml, specifies the genomes of interest and the aligner indexes to use. The config/fabricrc.txt file specifies details about the system and where to install the data.

The basic commandline is:

fab -f data_fabfile.py -H your_machine install_data_s3

and you can pass in custom biodata and fabricrc files with:

fab -f data_fabfile.py -H your_machine -c your_fabricrc.txt install_data_s3:your_biodata.yaml

In addition to downloading and preparing the data, the script will integrate these files with a Galaxy instance by updating appropriate Galaxy configuration files. This makes it useful for installing data to a local or cloud-based Galaxy server.

Not all of the genomes are hosted on the S3 bucket, but are still supported. If your genome fails to install with install_data_s3, you might be able to download the genome from Ensembl, etc and prepare it:

fab -f data_fabfile.py -H your_machine -c your_fabricrc.txt install_data:your_biodata.yaml

Using pre-built cloud images

Amazon

See the 'Getting Started with CloudBioLinux' guide on the CloudBioLinux website for a detailed description. The short version for users familiar with Amazon is:

  • Login to the Amazon EC2 console.
  • Click Launch Instance, and choose the latest CloudBioLinux AMI from the website in the community AMI section (search for 'CloudBioLinux').
  • After launching the instance, find the host details of your running instance from the Instances section.
  • Connect to your machine via ssh or VNC (using the Amazon PEM keys)

Supported environments

Docker

Docker provides lightweight local containers for Linux machines, allowing isolation without the associated overhead of full virtual machines. Include any of the standard CloudBioLinux commands inside a Dockerfile to use CloudBioLinux to build up the set of tools on your instance. See the Dockerfile examples for information how to write Dockerfiles.

To use a pre-built Docker image made with CloudBioLinux infrastructure, using this bcbio-nextgen Dockerfile, you can import the bcbio-nextgen container into your local docker environment:

docker import https://s3.amazonaws.com/bcbio_nextgen/bcbio-nextgen-docker-image.gz chapmanb/bcbio-nextgen-cbl

Amazon

A bare Linux image launched in Amazon EC2 is configured from another machine, i.e. your local desktop, using ssh and cloudbiolinux. See the Installation section for installing CloudBioLinux with fabric.

Any cloudbiolinux distribution can be used, including Ubuntu, Debian Linux and CentOS. We recommend using m1.medium or better instance for building a CloudBioLinux image from scratch, due to resource usage while compiling software.

  1. Go to the cloudbiolinux source and edit the config/fabricrc.txt, to match the system you plan to install on. Specifically, distribution and dist_name parameters specify details about the type of target.
  2. Start an Amazon EC2 base instance and retrieve it's DNS hostname:
  1. From your local machine, have CloudBioLinux install your Amazon instance:

    fab -f fabfile.py -H hostname -u username -i private_key_file install_biolinux
    
  2. When finished, use the Amazon console to create an AMI. Thereafter make it public so it can be used by others.

Vagrant and VirtualBox

Vagrant allows easy deploying and connecting to VirtualBox images. The setup is ideal for runnig CloudBioLinux on a desktop computer. Install VirtualBox and vagrant.

See the VirtualBox and Vagrant documentation for details on creating a local virtual machine from scratch with CloudBioLinux.

Through Vagrant additional facilities are available, such as a shared network drive. It is also possible to tweak the image (e.g. RAM/CPU settings, and getting the all important guest additions) by firing up virtualbox itself. For more information, see the documentation on the Vagrant website.

OpenStack/XEN/KVM/Eucalyptus private Cloud

As long as there is an 'ssh' entry to an running VM, CloudBioLinux can install itself.

For more on private Cloud and CloudBioLinux see ./doc/private_cloud.md.

EC2 quickstart

This provides a quick cheat sheet of commands for getting up and running on EC2 using Amazon's command line tools.

Initial set up

The first time using EC2, you'll need to install the toolkit and credentials for connecting on your local machine, following the getting started guide.

Login to your Amazon EC2 account and go to Security Credentials/X.509. Create a new certificate and download the public cert-*.pem and private pk-*.pem files. Put these in ~.ec2.

Install the ec2 api tools, which require java.

Set up .zshrc/.bashrc:

export EC2_PRIVATE_KEY=~/.ec2/pk-UBH43XTAWVNQMIZRAV3RP5IIBAPBIFVP.pem
export EC2_CERT=~/.ec2/cert-UBH43XTAWVNQMIZRAV3RP5IIBAPBIFVP.pem
export AWS_ACCESS_KEY_ID=<your access key>
export AWS_SECRET_ACCESS_KEY=<your secret access key>

To test, you should be able to run the command:

% ec2-describe-regions

Now generate a privatekey for logging in:

% ec2-add-keypair yourmachine-keypair

This will produce an RSA private key. You should copy and paste this to your .ec2 directory for future use:

% vim ~/.ec2/id-yourmachine.keypair
% chmod 600 ~/.ec2/id-yourmachine.keypair

Allow ssh and web access to your instances:

% ec2-authorize default -p 22
% ec2-authorize default -p 80

Starting an instance

Each time you'd like to use EC2, you need to create a remote instance to work with; the AWS console is useful for managing this process.

When building from scratch with Alestic images, you will need to increase the size of the root filesystem to fit all of the CloudBioLinux data and libraries. This is done by starting the instance from the commandline with:

% ec2-run-instances ami-1aad5273 -k kunkel-keypair -t m1.large
                    -b /dev/sda1=:20
% ec2-describe-instances i-0ca39764

On Ubuntu 10.04, you then need to ssh into the instance and resize the filesystem with:

% sudo resize2fs /dev/sda1

On 11.04 the resize happens automatically and this is not required.

Testing

BioLinux comes with an integration testing frame work - currently based on Vagrant. Try:

cd test
./testing_vagrant --help

Target VMs can be listed with

./testing_vagrant --list

Build a minimal VM

./testing_vagrant Minimal

Documentation

Additional documentation can be found in the ./doc directory in the BioLinux source tree.

LICENSE

The code is freely available under the MIT license.